[Jawed]

I think this is the first absolute confirmation of double-sided 3-3-2 memory configuration - not surprising, but still comforting to see.

Jawed

[DemoCoder]

To me, if you launch a card that is the absolute performance leader, or even on-par with the competition and sell it significantly cheaper, you made an idiotic business decision, because unless AMD has amazingly cheap manufacturing, and world-beating yields, then their costs are about the same. You sell your card very cheap when you don't have world beating performance and hope to take marketshare via price war, while suffering lower margins. But selling a world-beating performance card for less than the slower cards? I don't think AMD is this stupid and they are not in the business of Charity Giving, so either the price rumors are wrong, or the performance isn't that spectacular.

[NocturnDragon]

Originally Posted by DemoCoder

To me, if you launch a card that is the absolute performance leader, or even on-par with the competition and sell it significantly cheaper, you made an idiotic business decision, because unless AMD has amazingly cheap manufacturing, and world-beating yields, then their costs are about the same. You sell your card very cheap when you don't have world beating performance and hope to take marketshare via price war, while suffering lower margins. But selling a world-beating performance card for less than the slower cards? I don't think AMD is this stupid and they are not in the business of Charity Giving, so either the price rumors are wrong, or the performance isn't that spectacular.

What if the 400$ are after a $150 coupon for a high spec AMD cpu and a $50 coupon for a AMD brand new MB? You've heard it first! :P And I just made it all up, but it could make sense somehow. ...

[Geeforcer]

Originally Posted by DemoCoder

I don't think AMD is this stupid and they are not in the business of Charity Giving, so either the price rumors are wrong, or the performance isn't that spectacular.

Looking at their strategy over the last year, I would not be so sure.

[Silent_Buddha]

IF OEMs actually are buying R600's in the millions.

And that's a pretty plausible "if" considering Apple appears committed to the chip which would indicate other OEMs would also be interested especially if R600 drivers really are as stable as AMD indicates. And OEM's more than anything want stable drivers to go along with hardware, thus their aversion to using G80 in high volume product lines.

Actually, the thought of "millions" of R600's going to OEM's does make sense. It solves a few problems.

1. Chips with higher than desirable heat output are relegated to a market that can afford to have rather large and exotic heatsinks capable of dissipating said heat.

2. It avoids the situation that ATI had with the launch of R420 where OEMs got large initial supplies but neither OEM's nor consumer's were able to get enough to satisfy demand due to splitting allocation between the two.

3. Large sales to OEMs is basically money in the bank. There is no gamble about whether enough will sell into the retail channel to regain lost expenditure. At this point the burden is on the OEM to sell machines with the parts, AMD already has the money.

4. Filling demand from OEMs first allows them to then stockpile parts for retail without having to worry as much about try to balance retail availability vs OEM availability. Related to point 2.

5. Possibly giving more time to manufacture 65nm R650 such that it may launch as the high end part rather than a "too hot" R600 part. I give this VERY low probability.

6. Of higher probability it would also allow them to weed out and stock up on cooler running parts for the retail market.

Being a pragmatic optimist I'll tend to take a company at its word until such time as they are proved to be wrong.

As such, this would also help to explain the further delay, when AMD states they could have launched R600 into the retail channel a month or more ago.

As to secrecy on the OEM's parts. Considering most are build to order machines, I don't imagine it takes a lot of lead time to install new graphics cards into existing machines. I wouldn't be surprised if, due to AMD's paranoid secrecy on R600 that they are required to store purchases in a warehouse until closer to official launch. With added price incentives to make sure they comply.

Also, if true, it would indicate a couple possible things.

1. R600 drivers are significantly more stable and problem free than G80 drivers. Thus, OEM's will continue to forgo offering G80s and just move right on to R600s.

2. R600 really IS priced significantly lower than the competition, thus again, motivating OEMs to bypass G80 in favor of R600.

Anyways, all that is just a running train of thought on a rumor that may or may not be true.

We'll all find out in a few weeks a little more of the truth behind things I'd suspect.

Regards,
SB

[Silent_Buddha]

Originally Posted by DemoCoder

To me, if you launch a card that is the absolute performance leader, or even on-par with the competition and sell it significantly cheaper, you made an idiotic business decision, because unless AMD has amazingly cheap manufacturing, and world-beating yields, then their costs are about the same. You sell your card very cheap when you don't have world beating performance and hope to take marketshare via price war, while suffering lower margins. But selling a world-beating performance card for less than the slower cards? I don't think AMD is this stupid and they are not in the business of Charity Giving, so either the price rumors are wrong, or the performance isn't that spectacular.

Except for one very important thing as it relates to AMD right now.

They are absolutely focused on maintaining and if possible increasing market share at this point.

One only has to look at the CPU side of things to see them sacrificing margin at the cost of trying to maintain market share.

Considering the momentum that Nvidia has and the marketshare they are gaining because of it, I'd imagine a 400 dollar price point would do a few things for AMD.

1. Regain some lost marketshare and stop Nvidia's momentum.

2. Hopefully (from AMD's point of view) make the few people that even know R600 is delayed forget about the delay or at least forgive them for it.

3. Generate good-will and massive word of mouth advertising from enthusiasts who would presumably jump on a chip with performance on par with a 550-600 dollar part from the competition.

4. Force Nvidia to respond in kind, thus eroding their margins.

Is it worth the loss of increased cash flow from a higher margin on the part? Noone will really know until it's done and people have had a few months to examine the impact of such a move.

It's a gamble pure and simple. It could either work or it could backfire. But in the business world you won't make significant strides unless you're willing to take the gambles that go with it.

Regards,
SB

[Razor1]

SB the r600 costs around ~350 bucks to make (1 gig version) and 200-250ish for the 512 version, they would have no margins if they are to price it so low. That doesn't sound realistic if the r600 performs higher then g80. AMD doesn't have much to go by right now and not cut into more profits.

[Deusp]

Originally Posted by DemoCoder

To me, if you launch a card that is the absolute performance leader, or even on-par with the competition and sell it significantly cheaper, you made an idiotic business decision, because unless AMD has amazingly cheap manufacturing, and world-beating yields, then their costs are about the same. You sell your card very cheap when you don't have world beating performance and hope to take marketshare via price war, while suffering lower margins. But selling a world-beating performance card for less than the slower cards? I don't think AMD is this stupid and they are not in the business of Charity Giving, so either the price rumors are wrong, or the performance isn't that spectacular.

They're also completely forgetting about game theory, namely that nVidia won't sit still against this. If they see AMD/ATI coming with faster chips at a lower price, then nVidia will cut their chips to even lower prices, even if it means losing money. The only thing AMD will get out of this is a price war, and nVidia right now has much better finances. If nVidia is smart, they'll recognize the CPU price war AMD is waging and force them to wage a simultaneous GPU price war, with the intention of bankrupting AMD. This is a risk AMD cannot take. It's obvious that either the price rumor is wrong or that performance of the R600 is not impressive.

[neliz]

Originally Posted by Razor1

SB the r600 costs around ~350 bucks to make (1 gig version) and 200-250ish for the 512 version, they would have no margins if they are to price it so low. That doesn't sound realistic if the r600 performs higher then g80. AMD doesn't have much to go by right now and not cut into more profits.

Ahem.. wasn't everyone talking about mindshare?
Their no-margin low-volume unit is taking care of the mouth-to-mouth advertisement and generating a positive mindset (look nV's $1000 part is only marginally faster than this $400 board) and reflect positive radiance on the high-margin high-volume products.

It looks like a risky investment from AMD, but if you don't have money, why don't you generate marketing from sales?

BTW, WHERE are the benchmarks eh?

[Deusp]

Originally Posted by Silent_Buddha

Except for one very important thing as it relates to AMD right now.

They are absolutely focused on maintaining and if possible increasing market share at this point.

I can't believe a company that's losing as much money as AMD is right now is thinking that.

One only has to look at the CPU side of things to see them sacrificing margin at the cost of trying to maintain market share.

That's because Intel has the better product on the market currently. If Intel cuts prices, AMD must follow suit or be blown away.

Considering the momentum that Nvidia has and the marketshare they are gaining because of it, I'd imagine a 400 dollar price point would do a few things for AMD.

1. Regain some lost marketshare and stop Nvidia's momentum.

2. Hopefully (from AMD's point of view) make the few people that even know R600 is delayed forget about the delay or at least forgive them for it.

3. Generate good-will and massive word of mouth advertising from enthusiasts who would presumably jump on a chip with performance on par with a 550-600 dollar part from the competition.

4. Force Nvidia to respond in kind, thus eroding their margins.

Is it worth the loss of increased cash flow from a higher margin on the part? Noone will really know until it's done and people have had a few months to examine the impact of such a move.

It's a gamble pure and simple. It could either work or it could backfire. But in the business world you won't make significant strides unless you're willing to take the gambles that go with it.

Regards,
SB

Like I said, it will only force themselves into a price war with a company in a better financial state. It's a gamble that's almost certainly going to backfire and quite likely to put them out of business before they can see to its completion.

[Razor1]

Originally Posted by neliz

Ahem.. wasn't everyone talking about mindshare?
Their no-margin low-volume unit is taking care of the mouth-to-mouth advertisement and generating a positive mindset (look nV's $1000 part is only marginally faster than this $400 board) and reflect positive radiance on the high-margin high-volume products.

It looks like a risky investment from AMD, but if you don't have money, why don't you generate marketing from sales?

BTW, WHERE are the benchmarks eh?

what is the lowest margin high end card we have seen? The x1900xtx comes to mind, and they had a ok margins before the gf8's came out. How much marketshare is price going to garner if it forces people to buy new power supplies? Doesn't seem like a cost beneficial thing unless they start bundling PSU's with these things.

Is the x1900 series increasing mindshare, yes in the lower upper range, but only a few % points. 3-4%. Thats not enough to make much of a difference if they are going to drop their pants to lose 20-25% margins. At the top end its not even 3-4% its more like 2%-3%.

[neliz]

What if the low margin card, the 512MB GDDR3 version was on level with the GTX?

If you can run a GTX, you can run a X2900XT, there is like 20Watt theoretical difference between the two parts.

And the part that does sell for, $549 or whatever gives you the performance crown. you only need to pay the difference between the two versions for a potent PSU.

[Razor1]

thats very possible, I would think thats the route AMD is going to take.

[erick]

I think people are overlooking the fact that the high-end card does not necessarily need to make money. You can settle for just covering the costs involved, IF it helps you regain mindshare+marketshare and sell loads of cheaper, lower-performing cards.

Or better yet, if you already have sweet deals for those low-end parts (according to that TheInq article the RV610 is quite a hit with OEMs like Dell), then you won't risk anything more than temporarily losing some of your profit or, effectively, trading it in for more marketshare.

[CJ]

http://we.pcinlife.com/thread-747857-1-1.html

Pictures are already removed, but one supposedly showed a R600 scoring 11K at 1600x1200 in 3DM06.

Edit: fashionably late. same link as the one on NV News. :P

[DemoCoder]

Apple isn't in the business of buying "millions" of GPUs for Mac Pros, nor does Apple manufacture cards either. The Mac Pro isn't selling well, definately not in the millions.

The "R600 2900XT(X) is a loss leader to sell low/midrange" doesn't seem to be to be all that smart. Historically, IHV's have used the high-end cards to promote their leadership image to try and burnish their image with OEMs, but these cards typically have high margins and were sold at a premium, which paradoxically sometimes helps to increase one's image (Quadro anyone? 50 cent t-shirts sold for $50, etc)

The R600 isn't a console, where they expect to get licensing revenue from software by selling hardware at a loss. In fact, a user who buys an R600 "at a loss" isn't likely to buy a low/mid range ATI card too ala 'attachment rate' we are so fond of, so the high end card acts to cannibalize the sale.

At best, one could somehow hope that oodles of high end gamers with R600's would convince Joe-Blow to buy low end ATI cards. But if that's the strategy, why not just throw a LAN Party event and hand out a few thousand cards, or sell them at a loss *at the LAN event* for the biggest media "oomph"

I don't buy it. AMD desparately needs to increase its margins.

[Unknown Soldier]

Originally Posted by Kaotik

Except the good PR image it gives when you show it on your hardware, I mean, how would it look if Ferrari would demonstrate their brand new revolutionary gearbox on a Mercedes rather than their own Ferrari?

Sorry, we aren't talking cars here. These are GPU's which will be installed on Intel Hardware whether AMD like it or not. Holding the GPU's(R600) back so that they could be displayed on AMD's Barcelona which won't even see the day of light(for customers) until another 3-4 months down the line and will most probably be displayed on Intel's new QX6800 the day it does get reviewed by 90-95% of the sites, just goes to show that either AMD are arrogant or that someone made a big mistake in March in regards to both the GPU and CPU. That said, if Barcelona still comes out using higher watts that the Intel Quad but are only able to be as fast as the Quad's, I expect most people will stick to the Intel's.

And for your info, Ferrari allow Spyker to use their engine and Williams has allowed Toyota to use their seemless gearbox.

US

[_xxx_]

Originally Posted by erick

I think people are overlooking the fact that the high-end card does not necessarily need to make money. You can settle for just covering the costs involved, IF it helps you regain mindshare+marketshare and sell loads of cheaper, lower-performing cards.

This is really becoming the new myth. Where do you people get that sort of stuff? You may launch a part with some loss, but that will be limited according to thorogh calculations and planning. It's not like "hey, let's put millions into this part and then sell it with a loss to make the masses like us" or such

[Frank]

How fast is the RSX compared to the 8800? Because the Xenos is doing a bit better than the RSX it seems. And it's pretty certain that the R600 will be a bit bigger Xenos. That only leaves the question if the RSX is closely related to the 8800. Which seems to be the case.

[bigtabs]

Originally Posted by _xxx_

This is really becoming the new myth. Where do you people get that sort of stuff? You may launch a part with some loss, but that will be limited according to thorogh calculations and planning. It's not like "hey, let's put millions into this part and then sell it with a loss to make the masses like us" or such

Funny how you quoted him, yet completely misrepresented what he said.

[bigtabs]

Originally Posted by Frank

How fast is the RSX compared to the 8800? Because the Xenos is doing a bit better than the RSX it seems. And it's pretty certain that the R600 will be a bit bigger Xenos. That only leaves the question if the RSX is closely related to the 8800. Which seems to be the case.

I think the RSX is more like a 7600GT no?

[Twinkie]

Originally Posted by bigtabs

I think the RSX is more like a 7600GT no?

Its more of a 90nm 7800GTX clocked at 550Mhz.

edit - with abit of modification.

[Frank]

Originally Posted by bigtabs

I think the RSX is more like a 7600GT no?

That's the big secret. It is assumed, yes. And that's about what we know about it.

[INKster]

Originally Posted by Frank

That's the big secret. It is assumed, yes. And that's about what we know about it.

It is a bit big for a 7600 GT, no ?

Since it (RSX) has close to 300 Million transistors and the 7600 GT (G73) only has ~170 (even taking into account the different metric of transistor count between TSMC's 90nm and Sony's own 90nm processes), you must wonder what are those extra units for, especially when we already know that some 20M (the Purevideo processor) are not present at all in Sony/Nvidia's RSX.

People assume RSX is a close relative of the 7600 GT simply because of an abstract number such as the width of its memory bus being the same (128bit).
There's more to it.
The 128bit 7600 GT has no problem beating a 6800 Ultra with its 256bit bus, for instance.

[Unknown Soldier]

Originally Posted by Razor1

http://www.nvnews.net/vbulletin/showthread.php?t=89589

some more pics

I don't see any audio connector(or HDMI for that matter) in those pics. Of course that is the OEM version which should be 80nm. Don't know how packed the retail version is gonna look like with that audio(or HDMI connector). So nothing really new to what has already been shown 50000 times on the net since January(maybe except the memory ratings and configuration).

US